NX9 spline "chain points"

[mike944]

the old (to be retired) spline function has chaining options to automatically collect points to create the spline.

The "Studio spline" function doesn't seem to have this functionality. If it does, i can't seem to find it. When you have hundreds, or thousands of points that you need to spline, how are you supposed to do it without clicking each point?

Mike

 

[Toost]

If you use the Fit Spline feature. ( The new Fit Spline) you can group the points and then select the group.
I think that this is the new method for multi point selection.

Regards,
Tomas

 

[cowski]

In addition to what Toost wrote, below is a link to a journal that I wrote a while ago trying to resurrect the 'rectangle selection' of points. I don't guarantee it to be bullet proof, but should be better than nothing.

thread561-320527

p.s. I hope that you are not actually creating a spline through hundreds of points. With splines, less is better; more points gives more opportunity for unwanted inflections and ripples.

www.nxjournaling.com

 

[BubbaK]

> (The new Fit Spline) you can group the points and then select the group.

Does the group figure out the right ordering for the points, and correctly pass this ordering along to the spline function?

 

[Toost]

BubbaK,
i got curious and played around a little. I loaded an old part ( a sub-section from a windmill blade) added a two "stray points" and deleted a few points close to the new ones.
total number of points in my example was 129.
The old , retiring, spline does allow for chaining, but as Cowski notes, Spline through points should not be used if the number of points exceeds say ~8-10.
Used on this example i got a ... result. degree 3 x 128 segments. Not. Smooth.
But the chaining is smart. It chains the way i intended.

The Newer Studio Spline doesn't have a chaining option , which is logical per the above.

The new Fit Curve can both use chaining and groups. When using the group method and selecting a random point somewhere in the middle,...select=group, i get the same result ( selection order) as when using the old spline.
When using the chaining method and selecting "as in the old spline, pick the first , outermost, point and the opposite outermost i again get the same result.
The new fit spline has the nice option to move / remove points from the fitting process plus that one can fit analytical Lines/ circles / ellipses instead of "only" splines.

Regards,
Tomas

 

[BubbaK]

I don't necessarily agree that you should always use "Fitting" and never use interpolation "Though Points" with a large number of points. A lot depends on where the points came from. If they came from digitizing, then they contain small errors, and a spline-through-points will wiggle all over the place. If the points came from an analysis program or a calculation, then you *do* want the spline to pass through them exactly, and it won't wiggle. I don't see a need to use hundreds of points, but a few dozen would be reasonable, and selecting a few dozen points one at a time is going to be a royal pain.

 

[Toost]

I would be very careful trusting analysis data. It might very well be 100 % correct, but still produce unusable models.
I have seen quite a few nasty examples of the latter.
Many simulations are done using finite elements, a method which relies on simplification of the original data. - the input is radically simplified, the results is radically simplified and the tolerances are large/accuracy low. So the result is very good, but the coordinates are not.
Then we must make sure not to use too many of these points in our work.

The attached image is one of these examples. It's a section analysis of a windmill blade. (A wind power turbine)
The particular face has 1157x102 patches which is way beyond usable. ( all downstream operations will fail)
The shape is, as seen, "opposite of smooth", inflections and wobbly curvature.
Note that all these points used are probably within their tolerance, but they do not build smooth Nurbs curves / surfaces.

What/why went wrong on this one ?
A new user got to much data with too low quality , from a Finite element flow simulation, and used it all with too tight tolerances.
This surface cannot be used for anything.

Cleaning that data using methods like Fit Spline would have helped.

Regards,
Tomas

 

[BubbaK]

@Toost -- your reasoning shows that there are some cases where fitting (rather than interpolation) is the better way to construct a spline.

Have you ever worked with aerodynamics guys who engineer airfoils in turbine blades or wings. They give you a string of points, and they ask you to construct a curve. If you tell them "I made a curve; it doesn't go through the points, but it's pretty close, and look how few segments it has". How do you think that would be received??

So, fitting is not always the right approach. There are cases where you want interpolation. In, in these cases, chain selection of points is useful.